[论文] RULER: Representation-Level Verification of Machine Unlearning

论文概要

研究领域: ML
作者: Georgina Cosma, Axel Finke
发布时间: 2026-05-28
arXiv: 2605.27569

中文摘要

机器遗忘旨在从已部署模型中移除特定训练记录的影响,而无需从头重训。当前验证协议在输出层面通过成员推理、保留准确率和遗忘集准确率来检验,但模型可能满足所有三项指标,却在中间表示中仍编码着应被遗忘的记录。本文提出了RULER--一组表示层验证指标。Oracle比较指标M2衡量遗忘集记录在表示空间中是否仍占据与"从未训练过它们"的模型相同的位置;无Oracle指标M4仅从未学习模型的内部相似性结构中检测残留,无需重训。实验表明,四种近似遗忘方法全部通过了输出层评估,但M2在12个条件中的10个检测到显著残留(p<0.05),且效应量随遗忘比例增加而增长。M4在表格、图像、临床文本和人脸识别等多场景下作为预遗忘诊断工具,能检测到现有方法均无法完全抹除的身份级记忆信号。

原文摘要

Machine unlearning aims to remove the influence of specific training records from a deployed model without retraining from scratch. Current protocols verify this at the output level through membership inference, retain accuracy, and forget-set accuracy, but a model can satisfy all three whilst still encoding forgotten records in its intermediate representations. We introduce RULER, a set of representation-level verification metrics. The oracle-comparative metric M2 measures whether forget-set records occupy the same representational position as in a model retrained without them. The oracle-free metric M4 detects residuals from the unlearned model's internal similarity structure alone, without retraining. Four approximate unlearning methods all pass output-level evaluation, yet under a linear mixed-effects model M2 detects significant residuals in 10 of 12 conditions (p<0.05), with effect...

自动采集于 2026-05-29

#论文 #arXiv #ML #小凯